In this paper we present a detailed study of the structures and morphologies of a sample of 1188 massive galaxies with M _ { * } \geq 10 ^ { 10 } M _ { \odot } between redshifts z = 1 - 3 within the Ultra Deep Survey ( UDS ) region of the Cosmic Assembly Near-infrared Deep Extragalactic Legacy Survey ( CANDELS ) field .
Using this sample we determine how galaxy structure and morphology evolve with time , and investigate the nature of galaxy structure at high redshift .
We visually classify our sample into disks , ellipticals and peculiar systems and correct for redshift effects on these classifications through simulations .
We find significant evolution in the fractions of galaxies at a given visual classification as a function of redshift .
The peculiar population is dominant at z > 2 with a substantial spheroid population , and a negligible disk population .
We compute the transition redshift , z _ { trans } , where the combined fraction of spheroidal and disk galaxies is equal to that of the peculiar population , as z _ { trans } = 1.86 \pm 0.62 for galaxies in our stellar mass range .
We find that this transition changes as a function of stellar mass , with Hubble-type galaxies becoming dominant at higher redshifts for higher mass galaxies ( z _ { trans } = 2.22 \pm 0.82 ) , than for the lower mass galaxies ( z _ { trans } = 1.73 \pm 0.57 ) .
Higher mass galaxies become morphologically settled before their lower mass counterparts , a form of morphological downsizing .
We furthermore compare our visual classifications with Sérsic index , the concentration , asymmetry and clumpiness ( CAS ) parameters , star formation rate and rest frame U - B colour .
We find links between the colour of a galaxy , its star formation rate and how extended or peculiar it appears .
Finally , we discuss the negligible z > 2 disk fraction based on visual morphologies and speculate that this is an effect of forming disks appearing peculiar through processes such as violent disk instabilities or mergers .
We conclude that to properly define and measure high redshift morphology and structure a new and more exact classification scheme is needed .